* PHYless ethernet switch MAC-MAC serdes connection
@ 2015-01-20 13:20 Vijay
2015-01-20 18:14 ` Florian Fainelli
0 siblings, 1 reply; 4+ messages in thread
From: Vijay @ 2015-01-20 13:20 UTC (permalink / raw)
To: netdev
Hello All,
I have a custom board with Marvell 88E6046 ethernet switch which is
connected directly to freescale P1010 processor.
+------------+
| | +------------+
| | SGMII | |----- p0
| eth0 |-------------| p9 |
| | | |----- p1
| | +------------+
| | 88e6086 (Switch)
| |
| |
| | SGMII +-------------+
| eth1 |-------------| |
+--------------+ +-------------+
FS P1010 88E1512 ( PHY)
My dts looks like this,
mii_bus0: mdio@24000 {
#address-cells = <1>;
#size-cells = <0>;
compatible = "fsl,etsec2-mdio";
reg = <0x24000 0x1000 0xb0030 0x4>;
/* No PHY on external MDIO for 88e6086 Switch,
PHYless direct connection*/
/*------------*/
/* External PHY 88E1512 for eth1 */
phy1: ethernet-phy@1 {
interrupt-parent = <&mpic>;
interrupts = <3 1>;
reg = <0x1>;
};
};
mdio@25000 {
#address-cells = <1>;
#size-cells = <0>;
compatible = "fsl,etsec2-tbi";
reg = <0x25000 0x1000 0xb1030 0x4>;
tbi0: tbi-phy@11 { /* TBI
needed by gianfar for fixed-link eth0 */
reg = <0x11>;
device_type = "tbi-phy";
};
};
mdio@26000 {
#address-cells = <1>;
#size-cells = <0>;
compatible = "fsl,etsec2-tbi";
reg = <0x26000 0x1000 0xb1030 0x4>;
tbi1: tbi-phy@12 {
reg = <0x12>;
device_type = "tbi-phy";
};
};
enet1: ethernet@b1000 { /*switch
connected here*/
#address-cells = <1>;
#size-cells = <1>;
device_type = "network";
model = "eTSEC";
compatible = "fsl,etsec2";
reg = <0xb1000 0x1000>;
interrupts = <35 2 36 2 40 2>;
fsl,num_rx_queues = <0x1>;
fsl,num_tx_queues = <0x1>;
local-mac-address = [ 00 00 00 00 00 00 ];
interrupt-parent = <&mpic>;
fixed-link = <0 1 1000 0 0>;
tbi-handle = <&tbi0>;
phy-mode = "sgmii";
queue-group@0 {
#address-cells = <1>;
#size-cells = <1>;
reg = <0xb1000 0x1000>;
rx-bit-map = <0xff>;
tx-bit-map = <0xff>;
interrupts = <35 2 36 2 40 2>;
};
};
enet2: ethernet@b2000 {
#address-cells = <1>;
#size-cells = <1>;
device_type = "network";
model = "eTSEC";
compatible = "fsl,etsec2";
fsl,num_rx_queues = <0x1>;
fsl,num_tx_queues = <0x1>;
local-mac-address = [ 00 00 00 00 00 00 ];
interrupt-parent = <&mpic>;
phy-handle = <&phy1>;
tbi-handle = <&tbi1>;
phy-connection-type = "sgmii";
ptimer-handle = < &ptp_timer >;
queue-group@0 {
#address-cells = <1>;
#size-cells = <1>;
reg = <0xb2000 0x1000>;
rx-bit-map = <0xff>;
tx-bit-map = <0xff>;
interrupts = <31 2 32 2 33 2>;
};
};
dsa@0 {
compatible = "marvell,dsa";
#address-cells = <2>;
#size-cells = <0>;
dsa,ethernet = <&enet1>;
dsa,mii-bus = <&mii_bus0>;
switch@0 {
#address-cells = <1>;
#size-cells = <0>;
reg = <31 0>; /* Switch at SMI Add 0x1f */
port@0 {
reg = <0>;
label = "lan1";
};
port@1 {
reg = <1>;
label = "lan2";
};
port@9 {
reg = <9>;
label = "cpu";
};
};
};
I have disabled auto-negotiation in gianfar driver for fixed-link eth0
and associated TBI-PHY, so that SGMII link at 1Gb/ps can be forced with
switch port 9 GMII MAC.
Linux DSA driver is able to detect and configure switch. It discover
switch at external MDIO bus, DSA configure 'eth0' as master and 'lan1'
and 'lan2' as slave port.
After assigning IPs to eth0 (Switch) and eth1 ( with external PHY) like,
ifconfig eth0 up
ifconfig lan1 xx.xx.xx.x1 up
ifconfig lan2 xx.xx.xx.x2 up
ifconfig eth1 xx.xx.xx.x3 up
When I ping from any other host on network, only lan1 is able to
respond, because it was configured first. If any other port is pinged it
will be able to respond
only if 'lan1 is connected to network'. If ethernet cable from lan1 is
removed, rest of ports will not be able to respond, though they will be
receiving packets but
their TX counter will not increase.
It seems whichever port is configured first ( lan1 here ) will be able
to transmit packets. If eth1 ( connected to Phy not switch) is
configured first then only this
will be able to respond. All other ports then depends on eth1.
I tried to move 'eth1' to another mdio bus ( mdio@25000 ) by modifying
dts file above. But it seems 'PHY' and 'Switch' will only be detected if
they are on
mdio@24000.
What could be the reason of such strange behaviour ?
Why only one interface is controlling the bus its either eth0/lan0/lan1
or eth1. Rx counter of every interface increments but only one of these
have Tx counter incremented. If I disable switch ports ( lan0/lan1),
strangely eth1 also does not receive any packets ( though its Rx counter
increments but Tx is 0). Is it happening because dsa@0 also controlling
phy1 on same mdio bus ?
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: PHYless ethernet switch MAC-MAC serdes connection
2015-01-20 13:20 PHYless ethernet switch MAC-MAC serdes connection Vijay
@ 2015-01-20 18:14 ` Florian Fainelli
2015-01-21 7:09 ` Vijay
0 siblings, 1 reply; 4+ messages in thread
From: Florian Fainelli @ 2015-01-20 18:14 UTC (permalink / raw)
To: Vijay, netdev
On 20/01/15 05:20, Vijay wrote:
> Hello All,
>
> I have a custom board with Marvell 88E6046 ethernet switch which is
> connected directly to freescale P1010 processor.
>
> +------------+
> | | +------------+
> | | SGMII | |----- p0
> | eth0 |-------------| p9 |
> | | | |----- p1
> | | +------------+
> | | 88e6086 (Switch)
> | |
> | |
> | | SGMII +-------------+
> | eth1 |-------------| |
> +--------------+ +-------------+
> FS P1010 88E1512 ( PHY)
>
> My dts looks like this,
>
> mii_bus0: mdio@24000 {
> #address-cells = <1>;
> #size-cells = <0>;
> compatible = "fsl,etsec2-mdio";
> reg = <0x24000 0x1000 0xb0030 0x4>;
>
> /* No PHY on external MDIO for 88e6086 Switch,
> PHYless direct connection*/
> /*------------*/
>
> /* External PHY 88E1512 for eth1 */
> phy1: ethernet-phy@1 {
> interrupt-parent = <&mpic>;
> interrupts = <3 1>;
> reg = <0x1>;
> };
>
> };
>
> mdio@25000 {
> #address-cells = <1>;
> #size-cells = <0>;
> compatible = "fsl,etsec2-tbi";
> reg = <0x25000 0x1000 0xb1030 0x4>;
>
> tbi0: tbi-phy@11 { /* TBI
> needed by gianfar for fixed-link eth0 */
> reg = <0x11>;
> device_type = "tbi-phy";
> };
> };
>
> mdio@26000 {
> #address-cells = <1>;
> #size-cells = <0>;
> compatible = "fsl,etsec2-tbi";
> reg = <0x26000 0x1000 0xb1030 0x4>;
>
> tbi1: tbi-phy@12 {
> reg = <0x12>;
> device_type = "tbi-phy";
> };
> };
>
>
> enet1: ethernet@b1000 { /*switch
> connected here*/
> #address-cells = <1>;
> #size-cells = <1>;
> device_type = "network";
> model = "eTSEC";
> compatible = "fsl,etsec2";
> reg = <0xb1000 0x1000>;
> interrupts = <35 2 36 2 40 2>;
> fsl,num_rx_queues = <0x1>;
> fsl,num_tx_queues = <0x1>;
> local-mac-address = [ 00 00 00 00 00 00 ];
> interrupt-parent = <&mpic>;
> fixed-link = <0 1 1000 0 0>;
> tbi-handle = <&tbi0>;
> phy-mode = "sgmii";
> queue-group@0 {
> #address-cells = <1>;
> #size-cells = <1>;
> reg = <0xb1000 0x1000>;
> rx-bit-map = <0xff>;
> tx-bit-map = <0xff>;
> interrupts = <35 2 36 2 40 2>;
> };
> };
>
> enet2: ethernet@b2000 {
> #address-cells = <1>;
> #size-cells = <1>;
> device_type = "network";
> model = "eTSEC";
> compatible = "fsl,etsec2";
> fsl,num_rx_queues = <0x1>;
> fsl,num_tx_queues = <0x1>;
> local-mac-address = [ 00 00 00 00 00 00 ];
> interrupt-parent = <&mpic>;
> phy-handle = <&phy1>;
> tbi-handle = <&tbi1>;
> phy-connection-type = "sgmii";
>
> ptimer-handle = < &ptp_timer >;
> queue-group@0 {
> #address-cells = <1>;
> #size-cells = <1>;
> reg = <0xb2000 0x1000>;
> rx-bit-map = <0xff>;
> tx-bit-map = <0xff>;
> interrupts = <31 2 32 2 33 2>;
> };
> };
>
> dsa@0 {
> compatible = "marvell,dsa";
> #address-cells = <2>;
> #size-cells = <0>;
> dsa,ethernet = <&enet1>;
> dsa,mii-bus = <&mii_bus0>;
>
> switch@0 {
> #address-cells = <1>;
> #size-cells = <0>;
> reg = <31 0>; /* Switch at SMI Add 0x1f */
>
> port@0 {
> reg = <0>;
> label = "lan1";
> };
>
> port@1 {
> reg = <1>;
> label = "lan2";
> };
>
> port@9 {
> reg = <9>;
> label = "cpu";
> };
> };
> };
>
>
> I have disabled auto-negotiation in gianfar driver for fixed-link eth0
> and associated TBI-PHY, so that SGMII link at 1Gb/ps can be forced with
> switch port 9 GMII MAC.
> Linux DSA driver is able to detect and configure switch. It discover
> switch at external MDIO bus, DSA configure 'eth0' as master and 'lan1'
> and 'lan2' as slave port.
> After assigning IPs to eth0 (Switch) and eth1 ( with external PHY) like,
> ifconfig eth0 up
> ifconfig lan1 xx.xx.xx.x1 up
> ifconfig lan2 xx.xx.xx.x2 up
>
> ifconfig eth1 xx.xx.xx.x3 up
>
> When I ping from any other host on network, only lan1 is able to
> respond, because it was configured first. If any other port is pinged it
> will be able to respond
> only if 'lan1 is connected to network'. If ethernet cable from lan1 is
> removed, rest of ports will not be able to respond, though they will be
> receiving packets but
> their TX counter will not increase.
>
> It seems whichever port is configured first ( lan1 here ) will be able
> to transmit packets. If eth1 ( connected to Phy not switch) is
> configured first then only this
> will be able to respond. All other ports then depends on eth1.
Do lan1 and lan2 interfaces set the RUNNING flag, which would indicate
that the switch driver has detected a carrier for these individual ports?
Is the DSA switch driver correctly polling for LAN ports link statuses?
>
> I tried to move 'eth1' to another mdio bus ( mdio@25000 ) by modifying
> dts file above. But it seems 'PHY' and 'Switch' will only be detected if
> they are on
> mdio@24000.
>
> What could be the reason of such strange behaviour ?
> Why only one interface is controlling the bus its either eth0/lan0/lan1
> or eth1. Rx counter of every interface increments but only one of these
> have Tx counter incremented. If I disable switch ports ( lan0/lan1),
> strangely eth1 also does not receive any packets ( though its Rx counter
> increments but Tx is 0). Is it happening because dsa@0 also controlling
> phy1 on same mdio bus ?
DSA registers a slave MDIO bus interface which calls into the "real"
MDIO bus hardware to talk to the switch. In your case, you have both an
external switch and an external PHY connected to the same MDIO bus,
which is fine.
Your DT looks sane, although I suspect that lan2's PHY address is 3, and
not 2, as address 0 is typically a special MDIO broadcast address, so
Port 0's PHY can be read at PHY addr 1, Port 1's PHY can be read at PHY
addr 2 etc... If you remove "lan2", does it work slightly better?
>
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
--
Florian
^ permalink raw reply [flat|nested] 4+ messages in thread* Re: PHYless ethernet switch MAC-MAC serdes connection
2015-01-20 18:14 ` Florian Fainelli
@ 2015-01-21 7:09 ` Vijay
2015-01-23 16:49 ` Vijay Katoch
0 siblings, 1 reply; 4+ messages in thread
From: Vijay @ 2015-01-21 7:09 UTC (permalink / raw)
To: netdev
Florian Fainelli <f.fainelli <at> gmail.com> writes:
>
> Do lan1 and lan2 interfaces set the RUNNING flag, which would indicate
> that the switch driver has detected a carrier for these individual
ports?
>
> Is the DSA switch driver correctly polling for LAN ports link
statuses?
>
Yes RUNNING flag is set.
Whenever I unplug/plug Ethernet cable. I see notification from dsa about
link up/down correctly.
>
> DSA registers a slave MDIO bus interface which calls into the "real"
> MDIO bus hardware to talk to the switch. In your case, you have both
an
> external switch and an external PHY connected to the same MDIO bus,
> which is fine.
>
> Your DT looks sane, although I suspect that lan2's PHY address is 3,
and
> not 2, as address 0 is typically a special MDIO broadcast address, so
> Port 0's PHY can be read at PHY addr 1, Port 1's PHY can be read at
PHY
> addr 2 etc... If you remove "lan2", does it work slightly better?
>
No, lan 1 and lan2 addresses are 0 and 1 respectively. I have logged
these values and these are printed on console as:
DSA: dsa.c: Init slave bus
DSA: slave.c: dsa_slave_mii_bus_init sw addr =0x1f , slave_mii_bus_id =
0x0
DSA: dsa.c: register created slave_mii_bus
Debug: drivers/net/phy/phy_device.c: Creating PHY-> phy_id = 0x1410c89,
addr = 0x0
Creating PHY-> AUTONEG_ENABLE for phy_id = 0x1410c89, addr = 0x0
Debug: drivers/net/phy/phy_device.c: Creating PHY-> phy_id = 0x1410c89,
addr = 0x1
Creating PHY-> AUTONEG_ENABLE for phy_id = 0x1410c89, addr = 0x1
libphy: dsa slave smi: probed
net lan1: PHY already attached
DSA: slave.c: mii_bus-> name = dsa slave smi , id =dsa-0:1f
net lan2: PHY already attached
DSA: slave.c: mii_bus-> name = dsa slave smi , id =dsa-0:1f
My eth0 fixed-link address is also 0 and eth1 PHY address is 1.
Removing lan2 is not improving the situation.
Even if I do not set up eth1. Then same issue is happening between lan1
and lan2. Whichever is configured first ( using ifconfig) is able to
transmit packet (Tx increments). And rest all ports ( lan2) are just
receiving (their Tx counters remain 0). Even if I ping to lan2 ip,
response will be received only when ethernet cable is plugged to lan1 (
because it will only be able transmit the response), It does not matter
if cable is plugged to lan2 port.
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: PHYless ethernet switch MAC-MAC serdes connection
2015-01-21 7:09 ` Vijay
@ 2015-01-23 16:49 ` Vijay Katoch
0 siblings, 0 replies; 4+ messages in thread
From: Vijay Katoch @ 2015-01-23 16:49 UTC (permalink / raw)
To: netdev
It seems this behaviour is normal when assigning ip address to 3 devices
and all of them are 'up'.
When I examined packets using wireshark, correct packets are received on
other host.
I resolved the issue of interface that is configured first controlling
transmission, by running ifplugd daemon. So if cable connected to port
which was configured first is unplugged, that interface will go down,
allowing any of the other to transmit.
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2015-01-23 16:49 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2015-01-20 13:20 PHYless ethernet switch MAC-MAC serdes connection Vijay
2015-01-20 18:14 ` Florian Fainelli
2015-01-21 7:09 ` Vijay
2015-01-23 16:49 ` Vijay Katoch
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).